Research On Ontology Reasoning Based On Description Logic In Semantic Web

In order to solve the questions in World Wide Web, such as the data to be numerous and jumbled, unable to search the useful information fast, maintains inconveniently and so on, that Tim Berner-Lee proposed the Semantic Web concept in 2000, received people's widespread attention. He defined its seven levels structure, elaborated in detail how to construct network frame pattern which including the semantic information richly and able to processing automatically by computer.In the recent ten years, people have done in-depth research to the related levels in the structure which about the knowledge expressing, the framework constructing, the inquiring and inferring, the realization applying and so on. These researches have made certain progress in Semantic Web's core levels. The most important job is that some standard language for certain levels have been proposed, such as XML 1.0, RDF and OWL (OWL-DL). These have laid the foundation for the next construction of Semantic Network with unified regulations and the research of resource sharing.For realizing the Web information sharing and exchanging in the Ontology Vocabulary level, people have introduced the"Ontology"concept, to define the formalized standard explanation of sharing concept model. It has provided one kind of very good guiding principle and design tool, for people to realize that the standardized resources expression, the machine's intelligent understanding and the processing of semantic information.Description logic as one kind of knowledge expression language which contains logic information, is mainly used to express the thing and the connection among them. It first defines the related concepts in application domain, then uses these concepts to express the relations or properties of this domain, thus serves the purpose to express the individuals and objects in this application domain.In numerous formalized methods of knowledge expression, Description logic attracts people's special attention in the past ten years or more mainly because that they have clearly the"model - theory"mechanism, very suitably to express the application domain through the concept taxonomy, and they provided very good inference service.Semantic Web uses Ontology as the information expression, to realize that the information classifying and the relations inferring among them, which enables it to become a distributional architecture that by constitution of massive machine understandable data. In this architecture, the relation between data is expressed by some terminologies, and these terminologies also form one kind of complex network connection. Computer can obtain the data signification through these terminologies, and be able to use logic to do inference in this kind of network connection.Because of the good knowledge expression and the inference mechanism of Description logic, using the reasonable kind of DLs to make the Ontology information expression, and carrying on accuracy examination as well as recessive information inference and inquiry, is an important way to realize the application of Description logic in Ontology reasoning. At the same time, this way has a vital significance for realizing the design of reasoning system based on Description logic.In this paper, we will design and realize an Ontology reasoning system prototype based on Description logic, and confirm the correctness to the read-in Ontology files, as well as instruct user to carry on the recessive information inquiry to the Ontology files.The concrete thoughts and related technologies are as follows:1. Core technology's algorithm research: This article has done in-depth research for the existing related techniques in Ontology reasoning, and discussed the core connection between each inference problem. The technology of Ontology reasoning mainly includes two aspects: TBox inference and ABox inference. The TBox inference is used in examination of the relations between terminology definitions, can mainly be divided into several problems: the concept Satisfiability examination, the concept Subsumption relation examination, the concept Equivalence relation examination and the concept Disjointness relation examination, all of them can be transformed to the Satisfiability relation examination;The ABox inference mainly includes: Consistency examination, instance examination, the concept realization and retrieval problems. The former two problems are used to exam the correctness of the Ontology, can also be reduced to the Subsumption examination problem; However, the latter two problems are used to do the inquiry and inference on the knowledge base, it radically can be simplified to the determination of Satisfiability. In summary, we conclude the basic problem of Ontology reasoning to the Satisfiability examination problem, and use Tableau algorithm based on ALC as the core method to realize. 2. Classification algorithm research. The concept Classification, as an important task in Ontology reasoning, its goal lies in ascertaining the inclusion relation between the concepts of terminology. Moreover, the above classification allows us to construct the terminologies in the form of contain level. This kind of level has contributed useful information to the connection of different concepts and it can be used to enhance other inference service.The main problem in concept Classification is the comparison between concepts for Subsumption. From the introduction above, each contrast process may transform as the Satisfiability determination question. Because the worst time complexity of the Satisfiability determination is NExpTime, though using the optimization techniques to improve the efficiency, the result is not obvious. Furthermore, if there are too many concepts contained in TBox, contrasts in turn inevitably cost too much time. Therefore, how to reduce the number of contrast about concepts Subsumption relation is the key to realize the Classification optimization.In this paper, we use some optimization technologies, such as demonstration definition, restructuring, non-inclusion relation examination based on individuals, propose an improved algorithm ERXM about concept Classification. This method constructs an entire relational graph among the concepts, and using the simplified thought to compare the connection relations among concepts, stores them in the entire relational graph, and then uses it to instruct Subsumption relation determination among concepts. Because this algorithm may obtain massive connection information among concepts for us, just like not contain, not subclassof, not superclassof and so on. We can use the logical judgment to reduce times which transfers the Tableau algorithm to exam the satisfiability, finally raising the efficiency of concept Classification. It is shown that, from the experiments, this algorithm in improving Classification efficiency is practical effective. Especially when the Ontology contains more individuals, the improvement effect is more obvious. The reason is that when the quantity of individual increases, it can offer more connection information among concepts for us.3. System implementation. For the study of existing reasoners, such as Pellet, Racer, FaCT++, it is not difficult to find that they all follow the same design patterns which is ontology parser combine knowledge base and inference engine. Among them, the ontology parser is used to realize structure analysis, resources combination and grammar conversion for the read-in ontology files. Knowledge base is used to store the concepts, properties and individuals in the parsed ontology files. Inference engine eventually verify the correctness of the information in the knowledge base and do the query by inference. It is the key part of the entire system design and implementation. We also follow this pattern, adding the user interface part, to realize an ontology reasoning system prototype based on DLs.We use the simple, ALC-based ontology files as test body to check the abilities of this system. For the more complex expressive ontology file, we modify its structure to make it suitable for the system. The results show that it is useful to improve the efficiency of classification in the file.In this paper, the main job is that raising an improved classification algorithm, doing an in-depth research for the techniques of ontology reasoning based on DLs, designing and implementing of a reasoning system prototype, giving out the comparison and test of relevant characteristics.The next step will focus on the expansion of system to deal with more complex DLs, and the introduction of rule-based reasoning mechanism.